Results 1 - 10
of
15,262
Table 1. Characteristics of terrain and image stereo pairs of the three test cases
2000
"... In PAGE 5: ... This also gave the chance to study the behaviour of the proposed model under two entirely different ground cover types observed in the island that are known to influence the accuracy of image matching. The first test site, Loutra, has steep slopes and heavy vegetation (see Table1 ). The expected height accuracy that can be obtained for this type of area is 15-20 meters (Trinder et.... In PAGE 5: ...his type of area is 15-20 meters (Trinder et. al., 1994). The second, Apothika, has medium to steep slopes and very little vegetation (see Table1 ). The expected height accuracy for this test site is 10 to 15 meters.... ..."
Cited by 2
Table 1 shows the computed residues from three stereo pairs of non-capped Santa images and three stereo pairs of capped Santa images in cases: without denoising and with wavelet db30 denoising. We see that that the accuracy of fundamental matrices of non-cappedSanta images is always improved by wavelet denoising. Especially, the improve- ment in computing pair of view 2 and view 3 is essential. And the accuracy of fundamental matrices of capped Santa images are mainly degraded after wavelet denoising.
"... In PAGE 4: ...14 0.96 Table1 . Residues are always decreased in non-capped Santa after wavelet denoising, while the residues are mainly increased in capped Santa after wavelet denoising; nc- Santa = non-capped Santa, and cSanta = capped Santa.... ..."
Tables 1-3 give annotations for the data set, including the type of ordnance present in the image, the rough location in the right image of the stereo pair, and notes describing the condition of the ordnance, the images that were culled from the set, and those selected as training images.
1998
Cited by 1
Table 3: Result for the correlation technique applied to the stereo pair in Figure 2.
2000
"... In PAGE 18: ... Figure 2 depicts the result for the best correlation window. Table3 shows the results obtained using our method for di erent parameter settings, and in Figure 3 we present the best result that we have obtained with our method. It has an average error of 0:2639.... In PAGE 18: ... We point out that our method computes the disparity in 100% of the image, and that the the error is signi cantly smaller than the one of the correlation technique. Table3 also shows that our method is rather robust concerning the choice of the parameters. In fact, in most cases the results are better than for the correlation method with optimal window size.... ..."
Table 3: Result for the correlation technique applied to the stereo pair in Figure 2.
"... In PAGE 18: ... Figure 2 depicts the result for the best correlation window. Table3 shows the results obtained using our method for di erent parameter settings, and in Figure 3 we present the best result that we have obtained with our method. It has an average error of 0:2639.... In PAGE 18: ... We point out that our method computes the disparity in 100% of the image, and that the the error is signi cantly smaller than the one of the correlation technique. Table3 also shows that our method is rather robust concerning the choice of the parameters. In fact, in most cases the results are better than for the correlation method with optimal window size.... ..."
Table 3: Result for the correlation technique applied to the stereo pair in Figure 2.
"... In PAGE 19: ... Figure 2 depicts the result for the best correlation window. Table3 shows the results obtained using our method for di#1Berent parameter settings, and in Figure 3 we present the best result that wehave obtained with our method. It has an average error of 0:2639.... In PAGE 19: ... We point out that our method computes the disparityin100#25 of the image, and that the the error is signi#1Ccantly smaller than the one of the correlation technique. Table3 also shows that our method is rather robust concerning the choice of the parameters. In fact, in most cases the results are better than for the correlation method with optimal window size.... ..."
Table 1: Recognition results on the ETH-80 image set. 25 images in each category are randomly picked for training. The recognition rate is computed based on the remaining 16 image in each category. The average recognition rate is BJBIB1 (II) NORB jittered-textured dataset LeCun et al. created a data set [13] in which there are 5 generic categories, namely, four-legged animals, human figures, airplanes, trucks, and cars. There is an extra cat- egory of background for training. There are 10 object in- stances for each category. 1,944 stereo pairs of images for each object instance are captured under different viewing
Table 1 Performance of the proposed method for various stereo image pairs
2004
"... In PAGE 9: ... Therefore, the two examined methods may be used together in order to transmit some clusters with better quality than others. Table1 shows the experimental results for the four tested images. The estimated PSNR values express the perform- ance of the stereo image pair for distinct bit rates.... ..."
Cited by 2
Table 8: Results of self-calibration performed on stereo pairs of the calibration grid and
"... In PAGE 33: ...mage. We matched the views using the algorithm from Section 5.3 #28an exception was done for images of the calibration grid, for whichwewere provided - by a calibration software - with links 3D-2D between a point on the grid and its image; given two such links associated to the same point 3D, we matched its two images 2D#29. Having employed results from o#1B-line calibration #28the constraint: distance camera-scene #15 2:5m had been ful#1Clled#29, we ran the self-calibration algorithm for each stereo pair, obtaining estimations for #0B, related to every considered zooming #28see Table8 #29.... ..."
Table 4.1: Timings measured for one client computer in our system. The ap- proximate range constraint (see Section 4.2.2.2) is employed. The two numbers for the timing correspond to different images acquired by one of the stereo pairs.
2005
Results 1 - 10
of
15,262